Inspired by the actual structure and mechanism of biological neurons, artificial neural networks (ANNs) were created and have become a relevant part of computational artificial intelligence [1]. ANNs are a set of non-linear algorithms that have been successfully implemented for countless applications as they can be trained to originate highly reliable mathematical models that are able to act as, for instance, classifiers or estimators. These algorithms have been used to carry out numerous tasks in many fields ranging from industry to economics, all the way to biomedicine. In this last context is where the present research is centered and, specifically, to aid in the analysis of the data that is produced during breath analysis for cancer diagnosis. Our research group, which excels in the design and optimization of ANN-based models, collaborates in two projects which are focused on the early diagnosis of cancer. These projects are LCaos, which is primarily centered on lung cancer, and Volgacore, which focuses on gastric cancer. One of the main goals of these projects is to reach early diagnosis of cancer through breath analysis, as it has been shown in the past that human metabolism or clinical state are somewhat reflected in the endogenous volatile organic compounds (VOCs) present in exhaled breath [2,3]. During these projects, different methodologies have been employed to extract the underlying information that breath contains, leading to large databases that require complex mathematical treatment. This is where ANNs come into play, as they have been used to interpret this information and create useful mathematical tools for various applications. For example, they have been successfully trained to perfectly identify and accurately estimate the concentration of a set of polar and non-polar VOCs (comparable to those present in breath samples), individually or in mixtures, at low concentrations which were measured using cross-reactive silicon nanowire field-effect transistor sensors [4]. On the other hand, the data resulting from a study carried out with proton transfer reaction-mass spectrometry, which was used to measure the exhaled breath samples of lung cancer patients and healthy controls, was modeled with ANNs to create a classifying tool to distinguish both groups (correct classification rates were over 90%). Finally, recent results indicate that using ANNs to model databases obtained from the breath analysis of different gastric cancer patients and high- and low-risk groups, using gold nanoparticle-based cross-reactive sensor arrays, is a suitable alternative as well, as a successful preliminary analysis has been carried out, with high classification rates.